pgvecto.rs VS Weaviate

Pgvector is very slow, seconds to 10's of seconds, on filter and order by queries. Its maintainers are working on this as you can see in this currently 83 comment long issue on Github and pgvector.rs has made improvements as you can see here, but it's messy. I strongly believe that you don't want to be fighting through these issues when adding semantic search to your product. It's going to be a long term, hard fought struggle to keep up with pgvector's updates here and continuously tune it.

To evaluate the performance metrics in comparison to the original vector approach, we conducted benchmarking using the dbpedia-entities-openai3-text-embedding-3-large-3072-1M dataset. The benchmark was performed on a Google Cloud virtual machine (VM) with specifications of n2-standard-8, which includes 8 virtual CPUs and 32GB of memory. We used pgvecto.rs v0.2.1 as the vector database.

Please check out our documentation for more details. We encourage you to try out pgvecto.rs, benchmark it against your workloads, and contribute your indexing innovations. Join our Discord community to connect with the developers and other users working to improve pgvecto.rs!

It's kind of a tradeoff. Performance is just one factor when choosing the vector database. In pgvecto.rs https://github.com/tensorchord/pgvecto.rs, we store the index separately from PostgreSQL's internal storage, unlike pgvector's approach. This enable us to get multi-threaded indexing, async indexing without blocking the insertion, and faster search speed comparing to pgvector.

I don't see any fundamental reason why the index in Postgres would be slower than a specialized vector database. The query pattern of the vector database is simply a point query using an index, similar to other queries in an OLTP system.

The only limitation I see is scalability. It's not easy to make PostgreSQL distributed, but solutions like Citus exist, making it still possible.

(I'm the author of pgvecto.rs)

Hi, we've solved the problem you mentioned! Please take a look on our open source postgres vector extension https://github.com/tensorchord/pgvecto.rs.

Our index building process is significantly faster than pgvector on hnsw because we can utilize all the cores, whereas pgvector can only use one core. And for the filter support, we do support pre-filtering, which will guarantee enough results no matter the condition is.

Hi,

In our previous post titled “Do we really need a specialized vector database?” on HN (https://news.ycombinator.com/item?id=37097004) we discussed the importance of using a Postgres-based solution for vector search. However, we acknowledged that existing Postgres vector extensions lack support for metadata filtering.

We are excited to announce that we have now addressed this limitation. We are proud to be the first (https://github.com/tensorchord/pgvecto.rs) to enable conditional filtering directly on HNSW indexes within Postgres. This breakthrough allows for efficient and effective metadata filtering in combination with vector search, eliminating the tradeoff previously associated with using Postgres for this purpose.

We invite you to explore our updated offering and experience the benefits of seamless metadata filtering within a Postgres-based vector search system.

Yeah, I think in many cases you just need a vector search lib, instead of a DB.

And in some other cases, you may want postgres vector extension e.g. https://github.com/tensorchord/pgvecto.rs instead of a specialized vector db.

Today, i'll like to share a technique that i recently discovered whilst trying to measure how long it takes to execute a method or function in .NET. I faced this challenge when working with the vector search component of Weaviate. Weaviate is open-source vector database state-of-the-art machine learning (ML) models to turn your data - text, images, and more - into a searchable vector database. It stores both objects and vectors, allowing for the combination of vector search with structured filtering. Thus, the performance of search is integral especially for a database.

Of course, Weaviate isn't without its problems. The TypeScript and Python clients aren't fully crystallized (both moved very recently moved to new major versions that deprecate the previous ones), documentation on running Weaviate on-prem without Docker is mysteriously lacking, and issues that have relatively simple fixes take a long time to get resolved.

Weaviate (https://weaviate.io/)| hiring for Engineering | Remote | Full-time

Weaviate is an AI-native vector database that helps customers with hybrid search, RAG, and generative feedback loops. check out the open-source project here: https://github.com/weaviate/weaviate . we were recently named to the Top 50 AI Startups list published by Forbes. we are hiring for a number of different roles but are highlighting:

Senior Fullstack Engineer - https://careers.weaviate.io/jobs/4608044-senior-full-stack-e... - Go experience required. Not afraid to work up the stack as needed

Research Engineer - https://careers.weaviate.io/jobs/4898801-research-engineer

QA Engineer - https://careers.weaviate.io/jobs/4602648-qa-engineer - looking for someone with testing the build of a database

Technical Trainer - https://careers.weaviate.io/jobs/4694275-technical-trainer - join our Developer Relations team to ensure new customers and users are

Feel free to reach out directly to me - rich AT weaviate.io if you have any questions!

all openings here: https://weaviate.io/company/careers#jobs

Weaviate (https://weaviate.io/)| hiring for Engineering + Marketing| Remote between UTC-5 and UTC+2|

Weaviate is an AI-native vector database that helps customers with hybrid search, RAG, and generative feedback loops. check out the open-source product here: https://github.com/weaviate/weaviate . we are hiring for a number of different roles but are highlighting:

Senior Fullstack Engineer - https://careers.weaviate.io/jobs/4608044-senior-full-stack-e... - Go experience required. Not afraid to work up the stack as needed

In this tutorial, I introduce Weaviate, an open-source vector database, with the thenlper/gte-base embedding model from Alibaba, through Hugging Face's transformers library.